Update parameters to match Generalized Random Forest better

[rflperry]

Reference issue

None

Type of change

Feature request

What does this implement/fix?

Adds additional parameters to Uncertainty Forest to better match the generalized random forest implementations in econml, the R grf, and the theory in general. Also parallelizes the learning of the decision trees in order to speed up fitting. The added parameters are as follows:

• max_features (the number of features considered at each split node. Was all features before but most random forests use less than that by default)
• poisson_sampler (technical detail to match the grf theory)
• n_jobs (parallelization)
• max_samples : (the number of samples subsampled for use in each tree. Was all samples before but conventional honest forests use 0.5*n_samples by default)